基于数据驱动的移动目标卫星任务规划

卫星任务规划是卫星地球观测的重要前提。传统的卫星任务规划主要针对固定地面目标，不能满足日益复杂的任务需求。针对移动目标的卫星观测任务，提出了一种基于数据驱动的移动目标卫星任务规划方法。该方法在大量的移动目标数据的基础上，通过改进的长短期记忆神经网络算法预测了目标的未来轨迹和位置信息，并通过约束满足型遗传算法规划了基于预测算法结果的移动目标卫星观测任务。鉴于移动目标观测中约束和任务冲突的复杂性，约束满足型遗传算法以条件形式将约束嵌入到遗传算法中，并在算法中特别设计了冲突消除算子以解决任务冲突问题。仿真结果证明了该方法在解决移动目标卫星任务规划问题上具有优良的效率，并获得了很高的观测精度。

Satellite Mission Planning for Moving Targets Observation via Data Driven Approach

Satellite mission planning is an important premise for earth observation. Traditional satellite mission planning is mainly aimed at fixed ground targets, which cannot meet the increasingly complex mission requirements. This paper considers the moving target observation, and puts forward a method of satellite mission planning for moving target via data driven approach. Based on a large amount of moving target data, this method forecasts the future track and position information of the moving target through a modified long short-term memory networks algorithm, and proposes a constraint satisfaction genetic algorithm to plan the missions of moving target observation based on the results of prediction algorithm. In view of the complexity of the constraint and task conflict in the moving target observation, mission planning algorithm embeds the constraints into the genetic algorithm through conditional forms, and a conflict resolution operator is designed in mission planning algorithm to resolve task conflicts. The simulation results demonstrate the efficiency of the method to solve the mission planning and get higher observation accuracy.